• Mon - Sat 8:00 - 6:30, Sunday - CLOSED

# Echolocating bats rely on an innate speed-of-sound reference Echolocating bats rely on an innate speed-of-sound reference

## Significance

Animals rely on their senses to survive and reproduce. Sensory systems are subject to a trade-off between the advantage of flexibility that often comes with a cost of a prolonged learning period and the advantage of innateness, which is less successful in dealing with altered environments. Most bat species rely on echolocation—emitting sound signals and analyzing the returning echoes. An object’s distance can be assessed using echolocation given a reference to the speed of sound. Since bats experience a range of speeds of sound, we tested whether the encoding of the speed of sound is innate or learned. We found that bats’ reference to the speed of sound is innate and that it is not flexible during adulthood.

## Abstract

Animals must encode fundamental physical relationships in their brains. A heron plunging its head underwater to skewer a fish must correct for light refraction, an archerfish shooting down an insect must “consider” gravity, and an echolocating bat that is attacking prey must account for the speed of sound in order to assess its distance. Do animals learn these relations or are they encoded innately and can they adjust them as adults are all open questions. We addressed this question by shifting the speed of sound and assessing the sensory behavior of a bat species that naturally experiences different speeds of sound. We found that both newborn pups and adults are unable to adjust to this shift, suggesting that the speed of sound is innately encoded in the bat brain. Moreover, our results suggest that bats encode the world in terms of time and do not translate time into distance. Our results shed light on the evolution of innate and flexible sensory perception.

Every organism must reliably sense its environment in order to survive and reproduce (1). Some sensory systems are innate and unalterable (2), allowing for efficient use even by naïve newborn animals (3⇓–5). Others require learning or experience-dependent development—usually during a critical period during ontogeny (6, 7), though sometimes retained through adulthood (8), allowing for adapting sensing to changing environments (9, 10). The ability to accurately estimate distances with sub-centimeter accuracy is a hallmark of bat echolocation (11⇓–13). Bats achieve this accuracy by means of delay-tuned neurons—neurons that are activated by specific call–echo time delays, supposedly encoding target distance (14⇓⇓⇓⇓–19), although it should be noted that some work suggests that the tuning width of delay-tuned neurons might not allow the accuracy that bats exhibit in delay perception (20). Though delay tuning has been shown to be (at least partially) innate at the neural level (21), this has never been tested behaviorally. Namely, when a newborn bat takes off for the first time, does its brain correctly translate time delays into distance?

Translating time into distance relies on a reference of the speed of sound (SOS). This physical characteristic of the environment is not as stable as it may seem. The SOS may change considerably due to various environmental factors such as humidity, altitude, and temperature (22). Bats (Chiroptera) are a specious and widely distributed order of highly mobile and long-lived animals. They therefore experience a range of SOSs (with more than 5% variation, see below) between species, among species, and even within the life of a single individual. We therefore speculated that the reference of the SOS may not be innate to allow for the environmentally dependent SOS experienced by each animal.

To test this, we examined the acquisition of the SOS reference by exposing neonatal bats to an increased SOS environment from birth (Materials and Methods). We reared two groups of bats from birth to independent flight in two flight chambers: six bats in normal air (henceforth: “air pups”) and five bats in a helium-enriched air environment (Heliox), where the speed of sound was 15% higher (henceforth: “Heliox pups”). Notably, Heliox pups were never active and did not echolocate in non-Heliox environment (Materials and Methods). This 15% shift is higher than the ecological range and was chosen because it is high enough to enable us to document behavioral changes but low enough so as to allow the bats to function (that is, to fly despite the change in air density). In order to feed, the bats had to fly to a target positioned 1.3 m away from their wooden slit roost. Once the bats learned to fly to the target independently (after ca. 9 wk), we first documented their echolocation in the environment where they were brought up, and we then moved them to the other treatment for testing (Materials and Methods). Because bats adjust their echolocation parameters to the distance of the target, before and during flight (23), we used their echolocation to assess the bats’ target range estimates. If the SOS reference is learned based on experience, the bats raised in Heliox should have learned a faster reference, so that when they flew in normal air, they would have perceived the target as farther than it really was. We also ran the same experiments on adult bats to test adult plasticity.

## Results

Bats adapt their echolocation according to the distance of their target (24⇓⇓⇓⇓⇓⇓–31). In a previous study (23), we identified two echolocation parameters that enabled us to assess the bats’ time/distance estimation of a target even before they take off. Specifically, bats produced calls of shorter duration (Dur) and shorter interstrobe intervals (ISI) prior to takeoff when the landing target was closer. Except for the distance of the target, nothing else was changed in those experiments, strongly suggesting that it was the target distance that affected echolocation (Fig. 1A and Materials and Methods). Here, we used these echolocation parameters to quantify the bats’ sensory estimations of target distance under different SOS conditions [the short duration (<1.2 ms, most <1 ms) of the calls ensured that the targets were never in the forward overlap zone and thus could be detected (32)]. If newborn pups learned the SOS reference based on experience, the echolocation parameters (Dur and ISI) of the Heliox pups should have been longer when moved to air, in comparison with the control bats, because they would perceive the 1.3 m target as being 15% farther. This was not the case. Based on their echolocation, both groups perceived the target as being closer in the Heliox environment (as expected), but both groups used the same echolocation parameters: there was no difference between the groups raised in the air and the Heliox environments, strongly implying an innate SOS reference (Fig. 1 B and C); mixed effect generalized linear model (GLM) with the rearing origin and treatment as fixed effects, and individual and flight number as random effects: there was a significant difference between the two treatments (Heliox vs. Air) – P < 0.001 (df = 754.4, t-ratio = −6.5); but no effect of the rearing origin – P = 0.62 (df = 8.6, t-ratio = 0.5) or of the interaction between rearing origin and treatment – P = 0.26, showing the response did not differ according to rearing origin. Analysis of the magnitude of the response also did not reveal a significant difference between Heliox or Air pups (Mann–Whitney U test on individual mean response magnitude, P = 0.65 [df = 1, U = 27]). Pairwise post hoc analysis on individual means showed the effect of treatment was significant in both origin groups; that is, both groups perceived the 1.3 m target in Heliox as closer than the 1.3 m target in Air (Air: df = 5, t = 3.5, P = 0.009; Heliox: df = 4, t = 4.7, P = 0.005).

Fig. 1.

Pups underestimate target distance in increased SOS. (A) Echolocation terminology and parameters for analysis: a schematic illustration of the first part of a typical echolocation sequence produced during a flight to the target, including the last calls produced on the perch prior to takeoff. Following Amichai et al. (23), we used the duration of the calls in the last strobe (group of calls) produced before takeoff, and the interstrobe interval (ISI), the time elapsed between the last strobe before takeoff and the takeoff strobe in order to assess the bats’ range perception. (B and C) Both air-raised pups and Heliox-raised pups underestimated target distance in Heliox when preparing for flight (based on their pretakeoff ISI values, *P < 0.01 for values in Heliox versus values in Air for both pup groups). The two groups used similar echolocation values regardless of the SOS environment in which they were born and learned to fly. (B) A comparison of normalized data (Feature scaling normalization, Materials and Methods) between all air-raised pups and all Heliox-raised pups. Data shown are the medians + 10th, 25th, 75th, and 90th percentiles. (C1) Individual data of “air pups” (medians + 25th and 75th percentiles). (C2) Individual data of “Heliox pups” (medians + 25th and 75th percentiles).

A full analysis of the echolocation approach behavior of two pups from each of the two rearing conditions revealed that they behaved identically along the flight, strengthening the conclusion that the SOS reference is innate. The echolocation approach sequence is characterized by a gradual decrease of call intervals, duration, and other parameters (23, 33). In both groups, when flying in 15% SOS, the bats showed abnormal echolocation behavior: they immediately began their flight in the final phase of the echolocation approach sequence (i.e., with very short intervals), showing that they misperceived the distance of the target from take-off and during the entire flight (Fig. 2 and SI Appendix, Fig. S1).

Fig. 2.

Pups exhibit sensory error during flight in Heliox. (A) A schematic depiction of typical sensorimotor approach sequence. In normal air (Top), sensory preparation for landing begins at a typical distance from target (usually 90 to 70 cm) with the emission of “buzz-I,” followed by “buzz-II” that ends in landing on the target. The body posture accurately corresponds to the sensory sequence. In flight 1 during the approach, the legs are extended, the body is erected, and the flight is slowed (2, 3) in preparation for landing on the target. These motor behaviors correctly correspond to both echolocation and distance to target in normal air (see numbers in Top), but they only correspond to the echolocation in Heliox (Bottom) in which both echolocation and body posture do not correspond well to distance to target. (B) Juvenile bats start the flight toward target (left side of each panel, 130 cm from target) in Heliox (green) with much lower echolocation intervals than in air (blue). This was true both for pups raised in Heliox (top two panels) and for pups raised in air (bottom two panels), suggesting an innate SOS reference. Notice the decreasing difference between Heliox and air values as the bats near the target and the absolute sensory error decreases. Data shown are median ± 95% confidence intervals.

This result also allowed separating Heliox-induced lift flight deficits from sensory deficits. If only flight was impaired (due to reduced lift caused by the change in air density), we would have expected to see the typical echolocation approach sequence. However, in reality, bats echolocated as if they were much closer to the target, that is, as if they misperceived the distance of the target due to the change in the SOS. Note that flight speed did not change in the different conditions (∼1.5 to 1.6 m/s), and thus, at any moment, the bats were in the same actual distance to the target but perceived a closer distance in Heliox. Moreover, if reduced lift affected echolocation during flight, we would have expected to see differences in echolocation parameters between hits and miss flights in Heliox, since in hits, the motor approach was similar to that in Air. We found no such difference (mixed-effects GLM with IPI as response, condition and distance to target as fixed effects, and individual and rearing origin as random effects. Condition and distance are both significant effects, respectively: t-ratio = −11.9/−22.7, P < 0.001/<0.001; individual and origin were not significant, respectively: P = 0.089/0.048). Toward the end of the flight, the Heliox and Air echolocation parameters eventually converged to the same values. This is not surprising since 1) the Heliox error is always relative (estimated in percent of the measurement), so the closer the bats are to the target, the smaller the error becomes, and 2) at the end of the approach, the bats reach their physiological minima for these echolocation parameters, making it harder to further reduce them (34). Our results thus suggest that even pups that always experienced an elevated speed of sound will encode a slower speed reference that is typical for normal air.

Once observing that the SOS reference is likely innately encoded in bat-pups, we next set to test whether adults are able to adjust it based on environmental changes. Notably, the SOS is not constant, and it can change due to various environmental factors including air humidity and density and, most importantly, ambient temperature (22). Various bat species experience substantial changes in SOS. A Tadarida brasiliensis bat ascending in summer from ground altitude up to over 3,000 m (35) will experience a drop of ∼5% in the SOS within minutes. Other species can experience even larger changes along the season. Pipistrellus kuhlii bats that roost in the Israeli “Aravah” desert occasionally forage in evenings with temperatures reaching 40 °C in summer and below 5 °C in winter, accounting for a change of ∼7% in the SOS, and other desert species might experience even larger temperature ranges.

We tested three different Heliox environments ranging from highly non-ecological (27%) to slightly beyond the ecological range (10%). In all cases, we observed no adult plasticity. Based on their echolocation parameters, all bats underestimated the distance of the target in all Heliox levels, suggesting that they did not adjust their reference to the new SOS (Fig. 3 B–E). There was also no evidence of sensory adaptation over time—the bats did not adjust their echolocation during the period they spent in Heliox (not daily and not after 3 to 6 d in 27% Heliox, Table 1 and SI Appendix, Fig. S2). A mixed-effect GLM with condition, day, flight number, and the interaction day*flight number set as fixed effects and individual as random effect found that only condition, namely Heliox/Air had a significant effect on the different echolocation parameters (ISI: P = 0.001, df = 211.1, t-ratio = −3.33; DUR: P = 0.004, df = 314.8, t-ratio = −2.92) or after up to 13 d in 10% Heliox (Table 2 and SI Appendix, Fig. S3). A mixed-effect GLM with condition, outcome, day, flight number, and day*flight number set as fixed effects and individual as random effect found that only condition had a significant effect on the different echolocation parameters (ISI: P < 0.001, df = 342.3, t-ratio = −5.75; DUR: P < 0.001, df = 439.1, t-ratio = −5.27).

Fig. 3.

Adult bats underestimated target distance in increased SOS. Pretakeoff echolocation parameters were shorter in increased SOS (A and B: 27% SOS; C and D: 10% SOS). In all panels, box-plots show medians + 10th, 25th, 75th, and 90th percentiles of all individuals pooled. In all individual panels, lines depict medians + 25th and 75th percentiles. (A) ISI immediately before takeoff were shorter when SOS was increased by 27%. (A1) Normalized Data (Materials and Methods). (A2) Individual bat values for the different treatments. (B) Echolocation signal duration immediately before takeoff was shorter when SOS was increased by 27%. Values were more similar to those emitted toward a target at 90 cm in air. (B1) Normalized data from all individuals. (B2) Individual bat values. (C) ISI immediately before takeoff were shorter when SOS was increased by 10%, as expected from induced ranging error. (C1) Data normalized for presentation. (C2) Individual bat values. (D) Echolocation signal duration immediately before takeoff was shorter when SOS was increased by 10%. (D1) Normalized data from all individuals. (D2) Individual bat values for the different treatments. The differences were statistically significant: 27% SOS: P < 0.001 for call duration; P < 0.001 for ISI; and for both: GLM with condition as fixed effect and individual and flight number as random effects. 10% SOS: P < 0.001 for call duration; P < 0.001 for ISI; and for both: GLM with condition as fixed effect and individual and flight number as random effects.

View this table:Table 1.

No sensory adaptation over time (27% SOS)

View this table:Table 2.

No sensory adaptation over time (10% SOS)

The use of a very large increase in the SOS (27%) allowed us to test our result quantitatively; namely, we predicted that the echolocation parameters in 27% SOS will be similar to those of flying in regular air to a target at 90 cm (the equivalent perceived distance). Indeed, when we flew bats to a target at 90 cm in air, the values of the echolocation parameters were similar to those in 27% SOS (Fig. 3 B and C). This quantifiable measurement of the bats’ errors thus confirmed their complete range misperception without any adjustment, not even partially. It is possible that longer periods in Heliox would have led to sensory adjustment, but we believe this is not the case for a few reasons: 1) While the total accumulated flight-to-target time amounts to a short time, the bats were active, echolocating, and moving around in Heliox in their relative night for several hours each night; 2) The pups spent more than 9 wk in Heliox and showed no adaptation; and 3) In nature, a bat may experience rapid shifts in SOS within minutes (e.g., T. brasiliensis ascending, see above), so if there was an adjustment mechanism, it would be expected to operate quickly.

The 10% SOS treatment allowed us to establish that bats’ inability to adjust their SOS reference was not a result of failing to fly to the target. In the 10% SOS environment, the bats landed successfully in ∼51% of the trials over a period of up to 13 d in comparison to only 3% in the 27% SOS treatment, which was probably a result of lift deficits. We categorized misses in Heliox to “category I” misses, extremely short flights probably without motor adjustments, and “category II” misses, evident motor adjustments to overcome loss of lift, with misses probably due to sensory error (Materials and Methods, SI Appendix, Table S1 and Movie S1). However, even with successful landing, bats did not adjust the SOS reference over time (SI Appendix, Fig. S4 and Table S2 and Table 2). In accordance, there was also no improvement in landing performance over time in this environment (SI Appendix, Fig. S3).

To gain more insight regarding possible echolocation adaptations, we compared echolocation in successful flights which ended in landing on the target (“hit”) with the nonsuccessful flights which ended in landing before the target (“misses”) in the 10% SOS environment. The pretakeoff echolocation parameters did not significantly differ between hits and misses, implying that motor and not sensory adaptation distinguished between failure and success (SI Appendix, Fig. S5, P = 0.154 for ISI, Mann–Whitney test, U = 8355.5; and P = 0.488 for Dur, Mann–Whitney test, U = 45332.5). Since echolocation parameters did not predict success or failure, we surmised that flight kinematic adjustments were responsible for successful flights. Indeed, when analyzing the flight kinematics of two bats, we found evidence that they adjusted their flight (and not sensing) in Heliox: their wing stroke amplitude was 20% larger than when flying in air (Materials and Methods and SI Appendix, Fig. S6 and Movie S1). Supporting our finding that bats adjusted flight kinematics, bats that returned to air after spending time in Heliox sometimes showed a motor after effect, generating too much lift and overshooting the target (Movie S2). We therefore could not use success as indicator of sensory adjustments and relied upon echolocation parameters alone.

We also examined the detailed echolocation dynamics along the approach flight of two adult individuals (bats 3 and 4) in 15% SOS and found that just like the pups, the adults started their approach immediately at its final phase, corroborating that they misperceived the target’s distance all the way (Fig. 4).

Fig. 4.

Adult bats react to echo–return timing and not to distance and exhibit sensory error during flight in Heliox. Sensory error in Heliox. The bats start the flight toward target (left side of each panel, 130 cm from target) in Heliox (green) with much lower parameter values than in air (blue). This was true for category II misses (bat overcame loss of lift but undershot target) for both bats and for hits for bat 3 (bat 4 did not land successfully). Notice the decreasing difference between Heliox and air values as the bats near the target and the absolute sensory error decreases. Data shown are median ± 95% confidence intervals.

## Discussion

Newborn echolocating pups used the same SOS reference regardless of the SOS they experienced and used the same echolocation parameters and the same echolocation approach dynamics as adult bats, suggesting an innate SOS reference. Thus, from their first flights, pups “know” how to adjust their echolocation based on the time delay between their emission and the reception of an echo. This finding is in line with the observation that pups do not require nearby adult bats to develop intact echolocation (we and others have raised pups to hunt successfully without their mothers).

Adult bats showed no flexibility in updating their SOS reference to the environment. This was evident from several behaviors: 1) When changing the SOS in the environment, pretakeoff echolocation parameters correlated to the pulse–echo time delay and not to the actual distance of the target, and this misperception was not adjusted even after spending days in an elevated SOS environment; 2) The echolocation parameters of successful landings in an increased SOS environment did not differ from those of the unsuccessful attempts, suggesting no sensory learning (at least within our measurement error); and 3) Landing success did not improve over time, even in nearly ecological SOS shifts with bats landing too close when the SOS was higher than normal. Using a visual metaphor, we could say that the bats “focused” their echolocation too close.

It is conceivable that since changes in SOS are coupled with temperature changes, the bats need a temperature cue to adjust SOS reference, which may explain our negative results. Our experimental design could not test this hypothesis, which therefore remains possible; however, we posit that such a mechanism is 1) less likely as it would require a very accurate estimate of ambient temperature even when flying fast and thus at high wind speed (which alters temperature estimation) and 2) not sufficiently accurate because temperature is not the only ambient parameter influencing the SOS.

The inability to adjust the SOS reference might have behavioral implication. Both pups and adults exhibited an ill-approach echolocation pattern when flying in Heliox. The typical bat approach behavior entails an accurately timed sequence of sensorimotor actions (33, 36⇓–38). This sequence includes increasing the echolocation rate and emitting groups of calls on the sensory side and rolling the body while slowing down and stretching the legs on the movement side. These sensorimotor actions must be performed precisely at the right time for a successful catch or landing. The analysis of the bats’ echolocation approach behavior revealed that they misperceived the distance of the target, performing the entire sensorimotor approach too early. As a result, the bats entered the terminal echolocation phase and rolled their body to land much too early (Fig. 2A). In fact, the bats started their flight in Heliox immediately with echolocation that is typical for the terminal part of the attack (Figs. 2B and 4). At this last stage of the attack [that often begins at a distance of around 1 to 2 m (23, 33, 36)], the bats typically slow down dramatically (39⇓⇓–42), so that in nature, such an early response would have sometimes resulted in a complete miss of the prey (Fig. 2A). This might explain the observation that insectivorous bats intercept prey with their (wing or tail) membrane which provides much slack and allows less precise localization than when intercepting with the mouth.

Interestingly, bats inability to adjust their SOS reference implies that they encode their world in terms of time and not space. Even after successfully landing dozens of times in the 15% Heliox environment, the bats regarded the object as being 7.5 ms away and not 1.3 m away, and they synchronized their echolocation approach signals according to the time of the echoes regardless of the actual distance to target. Our results also support a previously suggested hypothesis that the temporal information flow encoded by neuronal responses serves as an internal model for the external acoustic environment (43).

Why then do bats not exhibit sensory flexibility? We hypothesize that the need of a rapid ontogeny might have determined the innate nature of the SOS reference. Newborn pups of many bat species begin flying by 3 wk and must reach independence within less than 10 wk, including using echolocation to hunt for insects on the wing, or risk not surviving the winter (44⇓–46). This ecological constraint has probably driven the evolution of accelerated ontogeny of echolocation-related organs during gestation (47), and it might have driven the innate neural basis of time-delay processing in the bat’s brain as indicated by several neural studies (21, 48). Moreover, the cost associated with a learning neural system may be high.

Sensory systems must adapt to the physical principles that are relevant for the animal’s habitat and ecology. Archerfish that shoot water at targets above the surface and herons that hunt submersed prey from above the water must “know” the expected light refraction angle to hunt successfully. These angles can vary with environmental changes, depending for example on water salinity. In cases when environmental variation is negligible relative to the animal's necessary precision [like in the case of the archerfish (49)], it might be useful to evolve an innate neural representation of the environment, but when the physical parameters change rapidly and substantially, it should be beneficial to evolve mechanisms that allow compensation for the potential error (50, 51) and enable adaptations of sensory expectations (8).

The inability of many animals to dissociate a sensory stimulus from a response in an environment that is rapidly changing due to anthropogenic activity can have a catastrophic impact on the survival of the species (52⇓–54). Determining which traits and behaviors should be flexible and which innate is a fundamental evolutionary task. The ability to learn has clear benefits, but it comes with various costs such as postponing independence or reduced efficiency of task performance at early ages (55). From a physiological point of view, innate sensory perception is in many ways more surprising than flexibility as the brain has to be wired correctly, taking the risk of making mistakes without any ability to correct. Here, we show a sensory system that evolved innateness despite conditions that should arguably enhance flexibility. This highlights the complexity of selection pressures and their evolutionary outcomes.

## Materials and Methods

### Animals.

In total, 24 adult female (16 pregnant, eight postlactating) P. kuhlii bats were captured under permit from the Israel Nature and Parks Authority, permit no. 2011/38137. Procedures were carried out under permit of the Institutional Animal Care and Use Committee operating according to the Israel Health Ministry, permit no. L-11–043. Bats were housed in the experimental setup (see below), with food and water available ad libitum. The bats were kept under a 16:8h light:dark cycle, with a corresponding temperature cycle of 26:23 ± 2 °C. In total, 18 pups were born in captivity and remained with their mothers until independent flight and feeding (see below).

### Bats SOS Environment.

P. kuhlii in the Arava Desert of Israel experience a very wide range of foraging temperatures during their annual cycle. The temperature at sunset exceeds 39 °C several times during summer and may be lower than 1 °C before dusk during the winter (Data taken from Paran meteorological station, Israel Meteorological Service, https://ims.data.gov.il/ims). The species is active throughout the year (56, 57).

### Experiment Setup.

We trained bats to fly from a perch in an elongated flight chamber (50 × 50 × 200 cm3) and land on an elevated (25 cm), vertical target situated 90 to 130 cm away from the perch, while changing the SOS in the flight chamber using different concentrations of helium-enriched air (Heliox). An increase in SOS will result in echoes returning earlier from the target than they would in normal SOS. The length of the chamber represents a typical target-approach distance for vespertillionid bats (23, 33, 36), and a typical successful flight to a 130 cm target lasts ∼0.9 s.

The flight chamber served as the bats’ living facility and was equipped with a slit-style roost at one end from which the bats took-off for flights. Video was recorded using two high-speed infrared (IR) cameras (Optitrack s250e, NaturalPoint, Inc.) at a rate of 100 fps (these were only present during the 27% and 15% SOS trials, see below). Echolocation signals were recorded at a 250 kHz sampling rate using an omnidirectional ultrasonic microphone (Knowles with USGH preamplifier, Avisoft Bioacoustics, Germany) situated under the perch. Flights were identified in video, and takeoff was located in the synchronized audio recording. For a detailed description experimental setup and data extraction, we refer the reader to ref. 23, in which the same setup was used.

### SOS Control and Measurement.

To control and change SOS, we adapted the previously described experimental setup into an open-flow positive-pressure system in which we could control the amount of helium in the medium. Because of the low molecular weight of the helium atoms, sound propagates faster in helium than in air. It is impossible, however, to keep an animal in an all-helium environment due to the lack of oxygen. We therefore built a system in which helium-enriched air (Heliox) constantly flows through the chamber and the oxygen percentage never drops below 20%.

To increase SOS, air in the chamber was replaced by Heliox in the following manner: a 20%:80% oxygen:helium mixture from a cylinder flowed into the chamber at a rate of 3 L/min for 15 min. SOS was then measured (see below), and once the desired SOS was established, Heliox flow rate was reduced to ∼0.3 L/min, and external air was added through the same tube at a rate of ∼0.1 L/min using a fish tank air pump (Schego Optimal, Germany). The chamber was air tight except for a 3 mm diameter filling nozzle and a 3 mm diameter dump nozzle, both located in the bottom of opposite walls. This allowed positive pressure to build in the chamber, preventing outside air from entering the chamber, and allowed for CO2 to be removed. The location of the filling and dump nozzles at the bottom of the chamber combined with the positive pressure in the chamber ensured a uniform dispersal of heliox in the chamber. To validate that SOS did not change during the experiment, it was measured three times every day—before filling with Heliox, right after, and at the end of the day. SOS was measured using two ultrasonic microphones positioned ∼1 m apart inside the bat chamber, in line with an ultrasonic speaker which was placed ∼20 cm away from the first microphone. An ultrasonic click (duration: 0.15 ms, peak frequency 45 kHz) was played by the speaker positioned in the same line of the two microphones and was recorded with both microphones into one multichannel recording. The difference between the signal’s time of arrival at the two microphones (ΔTOA) was measured to assess SOS. Henceforth, a 10% increase in SOS will be termed “10% SOS,” 27% increase in SOS will be termed “27% SOS,” and so on. The SOS measurement speaker and microphones were located on the bottom of the chamber, thus ensuring our measurement was of the minimal SOS in the chamber, but the differences within the chamber were probably small.

Bats were only active and echolocated during their subjective night (between 10:00 and 17:00). This was verified by recording echolocation around the clock and verifying the lack of echolocation calls during respective daytime and by constant motion-triggered video surveillance allowing to detect whenever they moved. During the rest of the time they were roosting silently within a 15 mm wide slit in their perch. This is a typical roost structure for this species in nature as well. Even if the pups echolocated inside the roost without our detection, this behavior would probably be useless for SOS calibration as the proximity or the walls (<15 mm) would cause an extreme overlap between call and echo. We thus filled the tank with helium on a daily basis ca. 1 h before their subjective night to allow for cage ventilation.

The increased SOS had no effect or a negligible one on the resonant qualities of the bats’ vocal tract. In three bats, we found some increase in peak frequencies of the first and second harmonics that are consistent with clutter response and show the bats viewed the targets as closer in Heliox, but the energy distribution across the two harmonics was not altered proving that this was not a result of the Heliox (SI Appendix, Fig. S7). Even in the 27% SOS condition, however, the fundamental harmonic was still the one containing the most energy.

### Experimental Design.

#### Pups.

To investigate sensory plasticity during early postnatal ontogeny, we reared bats from birth to the age of independent flight in either control conditions (normal air, henceforth “air pups”) or treatment (15% Heliox, henceforth “Heliox pups”). A total of four identical experimental chambers (described above) were prepared: one was used to house a group of eight pregnant females in normal air; one was used to house a group of eight pregnant females in 15% Heliox; and the two additional chambers were used for simultaneous testing of pups, one in normal air and one in 15% Heliox while keeping the rest of the individuals in their respective original environment. The females were trained to land on the platform and eat from the plate. To make sure their captive diet was not lacking, we dusted the mealworms with commercial mineral powder for reptiles and added vitamin D for infants to the bats’ drinking water. No pup suffered from rickets which is associated with poor diet for lactating female bats in captivity.

When the pups reached the age of 3 wk, we started handfeeding them mealworms to supplement milk from their mother. Feeding was done in the food plate to facilitate learning of target for later experiments. Pups normally start flying around 30 d after birth (29), first clumsily and progressively more competently. We did not begin experiments until we verified each pup’s ability to eat from the plate and to fly. Once we determined the pup was ready to begin experiments, we transferred simultaneously one air pup and one Heliox pup to separate flight chambers, each containing the pup’s original condition, and left them for baseline testing (described above) for 5 to 7 d. After baseline measurements we changed the condition in the chamber, testing the air pup in 15% Heliox and the Heliox pup in normal air for 5 to 7 d. The pups were then returned to their “home” chamber, and another pair was tested. Overall, we experimented with six air pups and five Heliox pups (SI Appendix, Table S1D).

The experimental setup was the same as described above (i.e., target at 130 cm), with one difference: due to technical problems, the microphone used in 15% Heliox was situated under the takeoff perch while the microphone used in air was situated under the target platform. This difference does not affect the accuracy of our interval measurements (as those are measured form start of call to the start of the next call), but it does affect our duration measurements: duration was measured as the length of the signal between a 9 db intensity drop before and after peak intensity (23), making it difficult to measure duration in a comparative manner. We therefore did not use duration in the 15% SOS condition.

These results have been acquired during experiments done with eight adult individuals that flew in three treatments: 27% SOS, 15% SOS, and 10% SOS.

##### 27% SOS.

With the exception of one individual, all bats performed flights to targets at a distance of 130 cm and 90 cm from the perch in air and to a target at 130 cm in 27 ± 2% SOS which is equivalent to ∼95 cm in air. One individual did not perform flights to 90 cm as it was a pregnant female that gave birth before it could complete the full treatment program and was subsequently released with its pup. Each individual was required to complete at least 14 flights at each treatment; for a detailed table of total flights analyzed, see SI Appendix, Table S1A. Treatments were completed in one session with several exceptions: two individuals refused to fly at 27% SOS after 2 d, and those treatments were therefore done in two sessions each, separated by 1 d in normal air. Treatment order was randomized between individuals.

We chose to increase SOS by 27% as the main treatment to provide values beyond noise level of the pretakeoff parameters—call duration and interstrobe intervals—which we identified in previous work (23) as good indicators for the bat’s range estimation.

All target-directed flights in air resulted in successful landings. Nearly all of the flights in 27% SOS resulted in an unsuccessful landing, where the bat underestimated the location of the target. The only exception was individual 1 who managed to land four times in 27% SOS. This limited success created two problems” (1) There were not enough data to compare successful and unsuccessful flights in Heliox, and (2) The bats might have not had ample opportunities to learn that the SOS has changed. To this end, we ran the same experiment, but this time in a 10% SOS environment.

##### 10% SOS.

The procedure here was uniform for all individuals: target distance was 90 cm, treatments were Air 1 –10% SOS – Air 2, and this control resulted in enough landing events to allow learning (see SI Appendix, Table S1B). We added a second air period with a target at the same distance to allow examination of any spillover of the adaptation, in case the bats adapted; the order of the treatments was thus the same for all bats here. When returned to normal air after several days in Heliox, some of the bats maintained their helium echolocation parameters, at least for a limited period (SI Appendix, Fig. S8). Rather than showing adaptation to the new SOS (there was no adjustment in Heliox), this observation is mostly probably an adaptation to environmental clutter, strengthening the conclusion that the bats perceived shorter distances in Heliox and as a response continue to use an echolocation strategy that is suitable for these distances. Similar adjustments to environmental statistics have been shown in humans in the behavioral (8) and neural levels (58).

### Estimating the Bats' Sensory Perception.

The lower molecular weight and density of Heliox compared to air affects not only SOS but also the lift generated by the bat during flight. This undesired effect meant that a simple hit/miss scoring of flight and landing success was not suitable to assess sensory impairment, as it was impossible to separate the two effects—sensory and motor—on behavioral success.

We thus analyzed the echolocation signal parameters emitted by the bat before take-off—call duration (Dur) and interstrobe interval (ISI)—which we have shown (23) to reflect the bat’s perception of the distance of the target (Fig. 1A). In this previous study, we showed that changing the distance of the target alone (while keeping all the rest constant) changes the values of these parameters. Combined with the fact that this was immediately followed by a flight to the target confirmed the bats focused their echolocation at the target. The two target distances (130 cm and 90 cm) were also chosen based on results (23) which suggest that it is relatively easy to infer the bat’s range assessment of these distances based on its echolocation parameters (Fig. 1). The 27% increase in SOS aimed to create a treatment in which if the bat compensates, it should exhibit echolocation like in the 130 cm condition, and if it does not, it should exhibit parameters that are similar to the 90 cm condition.

### Estimating Bats’ Adjustments to Flight Kinematics to Cope with Heliox.

To quantify at least to some degree some of the motor adjustments used by the bats to overcome loss of lift in Heliox, we repeated the experimental procedure for individuals 3 and 4 in air and in 15% SOS and target distance of 130 cm and recorded the flights with two synchronized high-speed video (Optitrack s250e, NaturalPoint, Inc., at a rate of 100 fps—fast enough for adequate capture of wing position given a measured wingbeat frequency of ∼15 Hz). We used one front-view camera to estimate wing parameters and a second, synchronized, top-view camera to determine distance to target using visible distance markings on the transparent ceiling of the chamber. This calibration method was sufficiently accurate given the magnitude of the response that was considerably higher than inaccuracies the method could introduce. Only one individual succeeded in landing on the target, but for both, we identified two “miss” categories. “Category I” included flights in which the bat clearly did not adjust motor responses to the lessened lift and landed on the floor less than 50 cm from takeoff. “Category II” were flights in which the bat clearly overcame the loss of lift and covered the entire distance to the target, passing just underneath it (arguably due to sensory error). We analyzed successful flight in air and Heliox, as well as “category II” misses in Heliox (N in air: 54/24 flights, in Heliox: 14/24 flights for individuals ¾, respectively, SI Appendix, Table S1C). We quantified three parameters: the distance between the wing tip and the center of the body at the highest point of the wing stroke cycle, when the wing is fully stretched (beginning of downstroke) and the distance between the same two points at the lowest point of the wing stroke cycle, when the wing is partly tucked as preparation for upstroke (end of downstroke). We then estimated the wing stroke amplitude—the displacement of wing tip from highest to lowest point. We did this both for Heliox and air conditions (with the target at 130 cm). These distances were measured in pixels from images taken at the same location for air and Heliox (35 cm from takeoff for individual 3 and 25 cm for individual 4).

### Statistical Analyses.

Statistical analyses were done using Sigmaplot software (Systat Software, Inc.) and JMP software (SAS Institute). When possible, parametric tests were used, and when data distribution was not normal, the nonparametric alternatives were used. For data normalization, we used feature scaling normalization which scales the data (of each bat) between 0 and 1 according to the equation: $\chi \text{'}=x-min\left(x\right)max\left(x\right)-min\left(x\right)$, where χ is an original value and χ’ is the normalized value. Charts were prepared in Sigmaplot, JMP, and Excel software (Microsoft Corp.).

## Data Availability

All metadata used in this study were derived from audio and video recordings and are available in the supplementary materials in Dataset S1. Raw audio and video files are available upon request from the authors.

## Acknowledgments

We thank Mor Taub for assistance with figure design, Sahar Hajyahia for assistance with experiments, and Tal Raz for assistance with acoustic data extraction. E.A. was partially supported by The Alexander and Eva Lester Scholarship for Postdoctoral Fellow at Tel Aviv University and by the Ecology, Evolution, Environment and Society Graduate Program Scholarship for a Postdoctoral Research Associate at Dartmouth College.